OpenFold
Our core model, inspired by AlphaFold2, with full training and inference code for high-accuracy protein folding.
Models & Code
The OpenFold software ecosystem
OpenFold develops state-of-the-art, permissively licensed AI tools for protein and biomolecular structure prediction. The entire training and inference stack, along with training datasets, is released under the Apache 2.0 license, making atomic-accuracy modeling accessible in open source for both research and commercial use.
OpenFold3
An open-source cofolding system predicting 3D structures of biomolecular complexes from sequence and molecular inputs.
OpenFold-Multimer
Inspired by AlphaFold-Multimer, for modeling protein–protein interactions and multimeric complexes.
OpenFold-SoloSeq
Predicts protein structures equivalent to OpenFold with no Multiple Sequence Alignment input required.
Ecosystem
Built on OpenFold
Tools and research from across the community built using OpenFold.
AQAffinity
SandboxAQ, high-speed, structure-free protein–ligand binding affinity prediction built on OpenFold3.
ApherisFold
Apheris, federated protein structure prediction powered by OpenFold.
ESMFold
Meta, language-model-based protein structure prediction.
MLPerf
OpenFold serves as a benchmark in the MLPerf HPC training suite.
AlphaLink
Integrating crosslinking mass-spectrometry data into structure prediction.
AlphaFlow
Generating protein conformational ensembles with flow matching.
DEERFold
Steering structure prediction with distance-distribution restraints.